Skip to main content
  1. Projects/

AI Voice Agent Call Centre Platform

Table of Contents

Client — logistics technology provider, South Africa
Role — Solution architect and pre-sales lead
Status — CEO-level proposal with live demo

The problem
#

Inbound support volume was growing faster than the client could hire. Call handling was human-staffed, agent turnover was high, and after-hours coverage did not exist. The client wanted “an AI that answers the phone” — which, as usual, was a symptom description rather than a requirement.

Discovery narrowed it down to three things that actually mattered: response latency low enough to feel like conversation, answers grounded in the client’s own operational data rather than model priors, and a clean escalation path to a human without the caller repeating themselves.

Architecture
#

LayerChoiceRationale
TelephonySIP trunk into a real-time media serverKeeps carrier integration decoupled from agent logic
Real-time orchestrationLiveKit agentsBarge-in and interruption handling out of the box
Speech to textStreaming ASR with domain vocabulary boostingProduct SKUs and place names are where generic ASR fails
ReasoningLLM with tool callingTools for order lookup, booking status, escalation
Knowledge groundingRAG over client SOPs and product dataPrevents confident fabrication on policy questions
Text to speechStreaming neural TTSFirst-audio-byte latency is the perceptual bottleneck
HandoffWarm transfer with conversation summaryThe caller does not start over

The design constraint that drove everything was a sub-1.5 second voice-to-voice latency budget, allocated explicitly across each hop. Anything slower reads as a bad line and callers start talking over the agent.

What I owned
#

  • End-to-end architecture and latency budget allocation
  • Build-vs-buy analysis across managed voice-agent platforms versus a composed stack
  • Commercial model — per-minute unit economics against fully-loaded human agent cost, including the honest version where AI containment rates fall short of vendor claims
  • Live demo delivered in a CEO-level meeting
  • Risk register: ASR accuracy on accented English, hallucination exposure on policy questions, carrier-side failure modes

What I would flag to any client considering this
#

Containment rate is the whole business case, and it is the number vendors are most optimistic about. A pilot scoped to the top five intents will tell you more in three weeks than any RFP response will.

Related